skills/awslabs/aidlc-workflows/aidlc/Gen Agent Trust Hub

aidlc

Pass

Audited by Gen Agent Trust Hub on Nov 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill operates by invoking a dedicated orchestration tool (the engine binary) to handle all lifecycle routing, configuration changes, and state management. These invocations are performed via the platform's standard tool harness using pre-defined verbs like orchestrate, config, and log.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input, such as project descriptions and plan names, which could potentially contain malicious instructions. However, it implements strong security controls to mitigate these risks.
  • Ingestion points: User-provided descriptions in frontmatter arguments and freeform chat replies used for plan composition and question answering.
  • Boundary markers: The skill does not use specific prompt delimiters but relies on strict execution boundaries.
  • Capability inventory: Execution of the orchestration engine, file system writes to specific project directories, and delegation to specialized sub-agents via Task().
  • Sanitization: The skill utilizes an advanced sanitization strategy to prevent command injection. It mandates single-quoting all user input passed to the shell, uses specific bash/PowerShell escaping for single quotes, and automatically offloads any input containing dangerous shell characters (e.g., $, backticks, pipes, redirects) to a temporary file (.aidlc-engine/answer-text/answer.txt) so that the engine can read it safely without shell evaluation.
  • [DATA_EXPOSURE]: The skill maintains internal state, session data, and logs within the .aidlc-sessions/ and .aidlc-engine/ directories inside the project workspace. Access is restricted to these project-local paths.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 15, 2026, 11:22 AM